any-stack.md

Use in any stack

The machine is framework-independent. Adapters only supply executors and transport events/snapshots.

Server request

const result = await runAgent(machine, {
  input: await request.json(),
  executors
});

if (result.status === "idle") {
  await frameworkStore.put(id, result.persist());
}
return Response.json(result);

Resume request

const snapshot = await frameworkStore.get(id);

let event;
try {
  event = parseAgentEvent(machine, await request.json());
} catch (error) {
  return Response.json({ error: String(error) }, { status: 400 });
}

const result = await runAgent(machine, { snapshot, event, executors });

if (result.ignored) {
  return Response.json(
    { error: `'${result.ignored.type}' does not apply right now` },
    { status: 409 }
  );
}

Parse at the boundary; runAgent adds no validation of its own. An event the restored state does not handle is ignored, not an error.

Long-lived UI actor

Bind executors with provideExecutors, create a normal XState actor, and connect it using the framework's XState integration.

Examples

• Next.js: next-host
• Cloudflare: cloudflare-agent-host, cloudflare-workers-ai-host
• Mastra, LangChain, Flue: mastra-host, langchain-host, flue-host
• AI SDK and AG-UI streaming: ai-sdk-ui-stream, tanstack-ai-stream

Flue is covered by the flue-host example only; there is no separate guide for it in these docs. Eve has no bridge in this repository, so a machine running under Eve needs a host adapter you write yourself, following the same shape as the examples above.

Use each framework's own storage, retry, queue, and interruption semantics.

Related

• Use in an XState 4 or 5 app: isolating the agent when the app's own XState version cannot change.